e at SciVerse ScienceDirect

Electoral Studies 32 (2013) 512–516

Contents lists availabl

Electoral Studies

journal homepage: www.elsevier .com/locate/e lectstud

Voter perceptions of agenda power and attributionof responsibility for economic performanceq

Raymond Duch a,*, Randolph Stevenson b

aNuffield College Oxford, Oxford, United KingdombRice University, USA

a r t i c l e i n f o

Article history:Received 4 May 2013Accepted 4 May 2013

Keywords:Economic votingPerceptions of agenda power

q Working Paper Nuffield Centre for ExperimeAugust, 2012.* Corresponding author.

E-mail address: raymond.duch@nuffield.ox.ac.uk

0261-3794/$ – see front matter � 2013 Published bhttp://dx.doi.org/10.1016/j.electstud.2013.05.013

a b s t r a c t

In two recent experiments (one in the lab and one over the internet) concerning collectivedecision making we determined that individuals mainly assign responsibility to the de-cision maker with agenda power and with the largest vote share (Duch et al., 2012). Wefound rather weak evidence that responsibility is assigned to decision makers with vetopower or allocated proportional to weighted voting power. Our conjecture then is thatindividuals in our online experiment who recognized the importance of proposal power inthe embedded experiment will be those more likely to exercise an economic vote for theConservative PM Party (since they are the agenda setter in the governing coalition) and forthe opposition Labour Party. The conjecture is confirmed. Essentially, the data show thateconomic voting at the individual level is confined to individuals who understand thevalue of proposal power. This in turn suggests that the economic vote itself is motivated bya coherent attempt to punish or reward parties that actually deserve it in the specific sensethat they were mostly responsible for choosing the policies that were implemented.Further, the strong reliance on proposal power as the workhorse of this mechanism ofaccountability, tells us that simple heuristics can do a lot of the work that cold rationalityand complex calculation have done in much of the previous discussion of economic voting.

� 2013 Published by Elsevier Ltd.

1. Introduction

Most voters in democratic elections cast a vote for aparty that was either a member of an incumbent governingcoalition or, if they expected to govern, would have to join acoalition government after the election (Armstrong andDuch, 2010). Recent work demonstrates that voters inthese coalition systems appear to behave as rational votersshould – that is significant numbers of voters in coalitionalcontexts engage in “coalition-directed voting”, i.e. tacticalvoting for particular parties in order to try to bring apreferred coalition to power. Kedar (2005) or Bargsted and

ntal Social Sciences,

(R. Duch).

y Elsevier Ltd.

Kedar (2009), for example, find that voters in contexts withcoalition governments engage in compensational voting,i.e. certain voters vote for more extreme parties with thegoal of shifting the policy position of governing coalitionscloser to their ideal points. Based on data from 86 electionsurveys conducted in 22 countries, Duch et al. (2010) findthat in 75% of these surveys more than 50% of voters makecoalition-directed calculations.

All of these recent works on the coalition-directed votemake simplifying assumptions regarding responsibilityattribution which is critical for the coalition-directedvote. How do voters map the observed distribution ofresponsibility (i.e., seats won and cabinet positions held)into actual administrative responsibility within the cabi-net? Our answer to this question is that voters approximatethe fully rational calculus of a coalition-directed voteby employing responsibility attribution heuristics for

1 See the large literature trying to develop indexes of voting power,(Banzhaf, 1965; Shapley and Shubik, 1954; Straffin, 1978; Gelman et al.,2002, 2004; Heard and Swartz, 1999).

R. Duch, R. Stevenson / Electoral Studies 32 (2013) 512–516 513

individual parties making up the governing coalition (Duchet al., 2012). We identify the heuristics voters employ forattributing responsibility for individuals parties in coalitiongovernment by conducting experiments about collectivedecision making.

Individuals clearly favor agenda power as a heuristicfor attributing responsibility for members of a collectivedecision making body. Two quite different responsibilityattribution experiments – one conducted in the NuffieldCESS lab and the other conducted online by CESS with arepresentative sample of the UK population – producethis same result (Duch et al., 2012). The lab experimentdemonstrates that when individuals have the opportu-nity to hold individual decision makers responsible for agroup decision, they primarily attribute responsibility tothe proposer and the party with the largest weight.The second experiment was designed to confirm thatindividuals hold proposers responsible because theybelieve they actually have a disproportionate impact onthe outcome (rather than simply treating proposers as afocal point for an individual’s anger or enthusiasm for acollective decision). Results from this second experimentconfirm the two lab findings: individuals believe that theproposer and the largest party in fact have a dispropor-tionate impact on the outcome of collective decisionmaking.

The surprising result from these experiments is theimportance that voters place on agenda setting powerwhen they attribute responsibility for collective decisions.The experiment is a powerful tool for isolating the attri-bution heuristics that individuals deploy when confrontedwith collective decisions. And while the experiment isextremely useful for isolating the general heuristic it leavessome important questions unanswered: Are voters whorecognise the importance of proposal power more likely tohold policy makers accountable? And, secondly, are thosewho recognise the importance of proposal power morelikely to attribute responsibility to the individual party inthe coalition with proposal power? This is a classic limi-tation of lab experiments – they are powerful tools fortesting theory but can be limited with respect to externalvalidity (Morton andWilliams, 2009). External validity is ofinterest to us here because the agenda setting result fromthe lab should have important implications for how wespecify vote choice models in contexts with coalitiongovernments.

In order to address this challenge we supplemented thelab experiment with an online survey of a representativesample of the UK population. The online survey included anexperiment designed to identify responsibility attributionheuristics – essentially an extension of the lab experiment.Participants in the survey were also asked an extensivebattery of questions, including vote preference, thatallowed us to estimate a standard UK vote choice model(Duch and Stevenson, 2008). Accordingly we had twopieces of critical information: First, the incentive-compatible decisions made by the participants in theattribution responsibility experiment provided a behav-ioural measure of the extent to which individual re-spondents valued proposal power in attributingresponsibility for collective decisions. Secondly, we had all

the necessary information to model vote preference forparties in and out of the governing coalition. The empiricalresults reported here will focus on responsibility attribu-tion for the government’s economic performance.

Our experimental results suggest that individuals thatattribute responsibility to individual decision makers forcollective decisions favor proposal power as a heuristicfor deciding who to reward or punish. And given the na-ture of this heuristic we think this makes theoreticalsense. Compared to other possible cues voters could useto attribute responsibility for collective decisions, pro-posal power is a particularly useful heuristic because itpredicts outcomes well in the real world (i.e., the PrimeMinisters often get their way or some favorable compro-mise), is easily applied to contemporary politics (i.e.,simply knowing which party is PM), and is comparativelysimple in its form. Compare proposal power to, forexample, using voting weights (e.g., seat shares) to pre-dict policy influence: the impact of voting weight iscertainly not linear, is highly contingent on the specifics ofthe situation, and in multiparty democracies requiresvoters to learn and remember a great deal more infor-mation than simply who is the PM.1 Given our evidence ofthe centrality of proposal power to responsibility attri-bution we conjecture that economic voting at the indi-vidual level is confined to individuals who understand thevalue of proposal power.

We expect that agenda power is typically, althoughnot always, associated with the Prime Ministerial party.For most coalition policies it is the Prime Ministerialparty that is considered by the voters to commandagenda power. This certainly seems to be the case withrespect to managing the economy although we havepresented evidence suggesting that with respect to theeconomy voters may consider the party of the FinanceMinister to have agenda power (Duch and Stevenson,2008). In the case of the incumbent UK coalition gov-ernment both the Prime Minister and the Chancellor ofthe Exchequer were controlled by the Conservative Party– agenda power with respect to economic policies wouldseem to be unambiguously in the hands of the Conser-vative Party. Accordingly, for those who favor proposalpower in responsibility attribution, economic votingshould be particularly focused on the Conservative, asopposed to the junior partner in the coalition, i.e., theLib-Dem Party. If our conjecture is correct then we wouldexpect that those participants that favored the agendapower heuristic would focus all of their economic voteon the Conservative Party. Respondents who haveinternalised this agenda setting heuristics are morelikely to hold accountable the member of the governingcoalition with clear proposal power for economic policy.The attraction of having incorporated the experiment ina representative online survey is that we are able toestimate responsibility attribution for each of the majorparties.

050

100

150

Freq

uenc

y

-4 -2 0 2random slope for dummy variable indicating which DM was proposer

Fig. 1. Variation in the revealed importance of proposal power on re-spondents guesses about policy outcomes: raw coefficients. Bayes pre-dictions for the slope for each respondent from a random coefficient modelof the distance between respondents’ policy guess and the ideal point ofeach decision maker. Negative coefficients mean the respondent thoughtpolicy would me closer to a DM’s ideal point if that DM was the proposer.

5010

0

Freq

uenc

y

R. Duch, R. Stevenson / Electoral Studies 32 (2013) 512–516514

2. Internet survey results

The analysis we present here will be based on anInternet survey of 1004 UK respondents that was con-ducted by the Nuffield CESS July 1–7, 2011.2 The surveyincluded a responsibility attribution experiment alongwiththe standard set of election study questions.

The design of the internet survey experiment and adiscussion of the results are reported in detail in Duch et al.(2012) and in the Online Appendix to this article. In theexperiment, respondents were asked to guess the outcomeof three collective decisions that had been decided by theweighted vote of five decision makers (DMs) prior to thesurvey. The voting weight distributions varied across thethree treatments and respondents were compensated forthe accuracy of their guesses. For each respondent, for eachof these questions, we calculated the spatial distance be-tween the respondents guess about the collective decisionand the ideal points of each of the five DMs. This gives usfive data points for each respondent for each of the threequestions. Taken together, these data points contain infor-mation about the respondent’s beliefs about the relativeinfluence of different DMs (with different seat weights,positions, and agenda powers) had on the collective deci-sion. We used these responses to model the characteristicsof decision makers that mattered most to our respondentsguesses. We found that, on average, our respondents put agreat deal of weight on proposal powers (i.e., the guessedpolicy outcomes were much closer to the ideal point of theproposer) controlling for all the other characteristic of de-cision makers mentioned above.

Most importantly, since we are concerned here withcharacterizing differences across respondents in theimportance they place on proposal power in forecastingpolicy outcomes (i.e., the extent to which they understandthe importance of such power), we allowed the coefficienton the variable identifying who was the proposer to varyover individuals. The estimates from this model are re-ported in the Online Appendix (and a full explanation of thevariables included in the model is in Duch et al. (2012)).3

For our purposes, however, what is important is the vari-ation in the coefficient capturing how much respondentsweighted proposal power in their assessment of policyoutcomes (i.e., did they guess the policy would be closer to,or further from, the ideal point of the proposer). Fig. 1 givesa histogram of this variation.

Notice that these coefficients are mostly negative, asthey should be since a negative coefficient means therespondent put policy closer to a decision maker’s idealpoint if she was the proposer. Still there is substantialvariation with about one-third of the respondents givingessentially no weight to proposal power in their choices. Inorder to facilitate the use of these coefficients in an

2 The internet panel sample was provided by Survey Sampling Inter-national (SSI) – their panelists are compensated with SSI points which areexchanged for money. A complete description of the survey and samplecharacteristics is available from the authors.

3 Duch et al. (2012) estimated this exact model, with the exception ofthe random intercept and slope that we added for the current application.

interaction term (where having negative and positivevalues leads to cumbersome interpretations), we trans-formed these coefficients to a continuous 0-1 metric. Thismonotonic transformation retains the key characteristics ofthe data (and the results, if not the ease of interpretation,are robust to using either). The histogram of this trans-formed variable is shown in Fig. 2.

Thus, in the models below, respondents with a 1 on thisvariable are those whose policy prediction was mostheavily impacted (in the correct direction) by proposalpower (so these are individuals who most “recognize” thevalue of proposal power), while those toward zero eitherdid not use proposal power or used it nonsensically.

3. The economic voting model and test of the mainhypothesis

As explained above, our survey and the experimentsembedded within it provide both survey questions with

0

0 .2 .4 .6 .8 1

random slope for dummy variable indicating which DM was proposer, transformed

Fig. 2. Variation in the revealed importance of proposal power on re-spondents guesses about policy outcomes: transformed coefficients. A valueof 1 means the respondent put a great deal of weight on whether a DM wasthe proposer in guessing how close policy would be to a DM’s ideal point.

Table 1Multinomial logit model of vote choice in UK, 2011.

Variable Coefficient Z

PM equationEconomic perceptions 0.02 0.07Economic perceptions X value

of proposal power�1.15 �2.91

Value of proposal power 5.17 3.66Age group 0.42 8.17Education quartile 0.04 0.59Income quartile 0.20 3.38Female 0.28 2.21Union household �0.40 �2.32Left-right placement 0.51 11.62Constant �6.08 �5.47Junior partner equationEconomic perceptions 0.31 0.78Economic perceptions X value

of proposal power�1.05 �1.98

Value of proposer power 5.17 2.62Age group 0.08 1.19Education quartile 0.16 2.07Income quartile 0.05 0.63Female 0.52 3.32Union household �0.49 �2.14Left-right placement 0.22 4.29Constant �5.73 �3.65

Observations: 1956 (652 respondents � 3 treatments).

-.5-.4

-.3-.2

-.10

.1.2

.3.4

.5

Cha

nge

in P

roba

blity

of V

ote

for P

arty

whe

n Ec

onom

ic P

erce

ptio

ns W

orse

n

0 .2 .4 .6 .8 1Importance of Proposal Power in Respondent's

Guess about Policy Outcomes

Change in Support for PM (Conservatives)

-.5-.4

-.3-.2

-.10

.1.2

.3.4

.5

Cha

nge

in P

roba

blity

of V

ote

for P

arty

whe

n Ec

onom

ic P

erce

ptio

ns W

orse

n

0 .2 .4 .6 .8 1Importance of Proposal Power in Respondent's

Guess about Policy Outcomes

Change in Support for Opposition (Labour and others)

Fig. 3. How the importance of proposal power on respondents guesses about policycategory 3, education quartile 2, non-union member, who places themselves at 4 o

R. Duch, R. Stevenson / Electoral Studies 32 (2013) 512–516 515

which to estimate an economic voting model and acontinuous measure of the extent to which each respon-dent has demonstrated (over our three survey experi-ments) that he or she understands the value of proposalpower in majority rule decision making. In this section weuse these data to estimate an economic voting model inwhich we interact economic perceptions with our measureof the respondent’s demonstrated understanding of thevalue of proposal power. Since the government in the UK atthe time of our survey was a coalition, we estimate amultinomial logit in which respondents may indicatesupport for the Conservatives (the PM party), the LiberalDemocrats (the junior partner in the government), or oneof the opposition parties. The variable capturing economicperceptions is a five-category variable asking whether therespondent thought the economy over the last year hadgotten much better, better, stayed the same, gotten worse,or gotten much worse. This variable was interacted withour measure of the respondent’s demonstrated under-standing of the value of proposal power. Given this setup,we expect the effects of worsening retrospective economicperceptions on support for the various parties (or groups ofparties) to be negative for the Conservatives and LiberalDemocrats and positive for the opposition parties. How-ever, we expect the size of these effects to be small (or even

-.5-.4

-.3-.2

-.10

.1.2

.3.4

.5

Cha

nge

in P

roba

blity

of V

ote

for P

arty

whe

n Ec

onom

ic P

erce

ptio

ns W

orse

n

0 .2 .4 .6 .8 1Importance of Proposal Power in Respondent's

Guess about Policy Outcomes

Change in Support for Cabinet Partner (Lib Dem)

outcomes impacts estimates of economic voting. Graphs are for a woman agen the left–right scale.

R. Duch, R. Stevenson / Electoral Studies 32 (2013) 512–516516

zero) for individuals who do not understand the value ofproposal power and larger for those who do. In addition, ifwe accept the auxiliary assumption (discussed above) thatindividuals think of the PM as having significantly greateragenda setting powers than his cabinet partners, we wouldexpect these effects to be most apparent for the PMs partyand least apparent for his junior partner.

Table 1 provides the coefficients from the estimatedmodel, and we can see immediately that the interactionterm on economic perceptions is both highly significantand in the expected direction (likewise for cabinet part-ners). Further, the signs on the various control variablesall make sense with, in general, older, rightist, higherincome, non-union member households supporting theConservatives.

Substantive effects from such models are, however,difficult to interpret from coefficients, so Fig. 3 gives therelevant substantive effects for each type of party and foreach level of understanding/use of proposal power in thesurvey experiment. For example, the graph in the upper leftpanel of Fig. 3 shows that the estimated size of the eco-nomic vote (the change in probability of voting for theparty when economic perceptions worsen) gets larger (i.e.,more negative) as respondents better understand (and use)the distribution of proposal power over decision makers toguide their policy expectations. This holds true for cabinetpartners, but quite weakly which is consistent with ourassumption that voters consider the party of the PM to bethe chief agenda setter in cabinet. Finally, the graph forthe opposition is the mirror image of the other two (as itmust be).

As we pointed out earlier, the lab and online experi-ments on responsibility attribution singled out proposalpower as a critical heuristics that determines the individualdecision makers that are rewarded or punished for a col-lective decision. The decisions made by the participants inthe online experiment allowed us to recover a behaviouralmeasure of the extent to which individual respondentsfavor proposal power when they attribute responsibility forcollective decisions. As we would expect, this proposalpower metric varies across the 1008 British respondents inthe online survey. Given the centrality of the proposalpower heuristic in our experiments our intuition was thatthose who more strongly favor proposal power for attri-bution responsibility in the experiments would also bemore likely, in general, to attribute responsibility for thegovernment’s performance managing the economy. Theresults reported in this section clearly suggest this is thecase. We do not see a similar effect for the junior partner inthe coalition government (the Lib-Dems) which is what wewould expect – those scoring high on the proposal powermetric recognise the Conservative Party as having proposalpower over the economy and hence reserve their economicvote for them.

4. Conclusion

Overall, we find these results striking. Essentially, theyshow that economic voting at the individual level isconfined to individuals who understand the value of pro-posal power. This in turn suggest that the economic votingitself is motivated by a coherent attempt to punish orreward parties that actually deserve it in the specific sensethat they were mostly responsible for choosing the policiesthat were implemented. Further, the strong reliance onproposal power as the workhorse of this mechanism ofaccountability, tells us that simple heuristics can do a lot ofthe work that cold rationality and complex calculation havedone in much of the previous discussion of economicvoting (including much of our own work on the topic). Wehope then that this simple study will stimulate more workthat attempts to combine survey experiments and tradi-tional surveys in ways that help reveal the actual mecha-nisms bywhich voters achieve democratic accountability incomplex political systems.

Appendix A. Supplementary data

Supplementary data related to this article can be foundat http://dx.doi.org/10.1016/j.electstud.2013.05.013.

References

Armstrong, D., Duch, R., 2010. How voters anticipate post-election coali-tion formation. Electoral Studies 29 (3), 308–315.

Banzhaf, J.F., 1965. Weighted voting doesn’t work: a mathematical anal-ysis. Rutgers Law Review 19, 317–343.

Bargsted, M., Kedar, O., 2009. Coalition-targeted Duvergerian voting: howexpectations affect voter choice under proportional representation.American Journal of Political Science 53, 307–323.

Duch, R.M., May, J., Armstrong, D., 2010. Coalition-directed voting inmulti-party democracies. American Political Science Review 104 (4),698–719.

Duch, R.M., Stevenson, R., 2008. The Economic Vote: How Political andEconomic Institutions Condition Election Results. Cambridge Uni-versity Press, Cambridge.

Duch, R., Przepiorka, W., Stevenson, R., July 2012. Responsibility Attri-bution for Collective Decision Makers. Working Paper Nuffield Centrefor Experimental Social Sciences.

Gelman, A., Katz, J.N., Tuerlinckx, F., 2002. The mathematics and statisticsof voting power. Statistical Science 17, 420–435.

Gelman, A., Katz, J.N., Bafumi, J., 2004. Standard voting power indicesdon’t work: an empirical analysis. British Journal of Political Science34, 657–674.

Heard, A.D., Swartz, T.B., 1999. Extended voting measures. CanadianJournal of Statistics 27, 173–182.

Kedar, O., 2005. When moderate voters prefer extreme parties: policybalancing in parliamentary elections. American Political Science Re-view 99 (2), 185–199.

Morton, R., Williams, K., 2009. From Nature to the Lab: ExperimentalPolitical Science and the Study of Causality. Cambridge UniversityPress.

Shapley, L.S., Shubik, M., 1954. A method for evaluating the distribution ofpower in a committee system. American Political Science Review 48,787–792.

Straffin, P.D., 1978. Game Theory and Political Science Chapter ProbabilityModels for Power Indices. University Press, New York, pp. 477–510.